Sam Altman wants to recreate Bell Labs. So does President Donald Trump’s administration. And Mark Zuckerberg. And Jeff Bezos.

In recent weeks, a wave of announcements has revealed Silicon Valley’s newest obsession: building modern versions of the legendary corporate research institutions where scientists once pursued breakthrough discoveries on company payrolls. There’s Episteme, a San Francisco operation backed by Altman and SoftBank’s Masayoshi Son.

There’s Periodic Labs, which lured more than 20 researchers away from OpenAI, Meta, and Google with more than $300 million in funding to build AI systems that can run physical experiments. Anthropic launched Claude for Life Sciences to accelerate drug discovery research. Bezos just led a $106 million round for Profluent, a startup using AI to design proteins and gene-editing tools from scratch.

All of them invoke the same golden age. Bell Labs developed the transistor and the mathematics underlying all digital communication. Xerox PARC invented the graphical user interface. IBM’s Almaden Research Center pursued foundational science, insulated from quarterly earnings calls.

Even the White House wants in, announcing the Genesis Mission to direct the Department of Energy to unite national laboratories for AI-accelerated discovery.

But that announcement masks a broader retreat from science.

Trump is slashing the research funding that built the internet

The Trump administration’s proposed fiscal 2026 budget would slash federal research and development spending by 22% compared to last year, according to the American Association for the Advancement of Science. The National Science Foundation would lose half its funding.

Federal research funding has an extraordinary track record. The internet began as ARPANET, a Defense Department project. GPS started as a military navigation system. The Human Genome Project was a government-funded effort that laid the groundwork for modern genomics and personalized medicine. The mRNA technology behind COVID vaccines emerged from decades of publicly funded basic research at universities and the National Institutes of Health.

But federal funding for the fundamental science that leads to breakthrough discoveries has been declining for years, from 52% of the total in 2012 to 41% in 2023. The business sector has picked up some slack, growing from 21% to 35%, but that shift comes with strings attached. Business funding tends to focus on shorter-term applications and measurable returns, not the patient, open-ended research that leads to fundamental breakthroughs.

And even that business contribution represents a shadow of what corporations once invested in science. They were killed by shareholder value maximization, the management philosophy that demanded every division justify its existence through immediate profitability. Jack Welch at General Electric epitomized this thinking in the 1980s and 1990s and helped make it mainstream. Long-term research that might pay off in a decade became indefensible when quarterly earnings reports dominated decision-making.

In the 1960s, chemical giant DuPont published more papers in American Chemical Society journals than MIT and Caltech combined. By 2006, just 6% of R&D Magazine’s innovation awards went to Fortune 500 companies, down from 47% in 1975.

The same capitalist logic that made Altman, Bezos, and Zuckerberg successful is what made Bell Labs-style research impossible for AT&T, IBM, and Xerox: move fast, optimize for growth, maximize shareholder value.

An imperfect solution to a policy failure

Now these billionaires are trying to recreate what capitalism destroyed. The key difference is artificial intelligence.

Without AI, it’s unlikely any of these initiatives would exist. The implicit promise running through all these initiatives is that AI will make research faster, cheaper, and more productive, accelerating discoveries that might take decades down to months or years. It’s a compelling theory, but one with little proven track record at the scale these initiatives envision. What’s certain is that without this AI optimism, it’s hard to imagine venture capitalists and billionaires lining up to fund open-ended research.

But what’s the alternative? Washington isn’t coming to the rescue. Universities are caught in bureaucratic quicksand, with at least one estimate suggesting academic scientists spend just one-third of their time on actual research. And corporations have largely stopped doing basic science. The share of basic research in total business R&D fell from 30% in 1985 to less than 20% in 2015.

Relying on billionaire-backed research isn’t ideal. These initiatives exist not because they suddenly discovered the value of patient capital for basic research. They exist because federal research policy has collapsed, AI is having a moment — and even billionaires want to be remembered for more than optimizing ad delivery and selling dishwasher tablets.

📬 Sign up for the Daily Brief